Abstract
In many signal processing and data mining applications, we need to approximate a given matrix Y of “sensor measurements” over several experiments by a low-rank product Y ≈ A·X, where X contains source signals for each experiment, A contains source-sensor mixing coefficients, and both A and X are unknown. We assume that the only a-priori information available is that A must have zeros at certain positions; this constrains the source-sensor network connectivity pattern.
In general, different AX factorizations approximate a given Y equally well, so a fundamental question is how the connectivity restricts the solution space. We present a combinatorial characterization of uniqueness up to diagonal scaling, called structural identifiability of the model, using the concept of structural rank from combinatorial matrix theory.
Next, we define an optimization problem that arises in the need for efficient experimental design: to minimize the number of sensors while maintaining structural identifiability. We prove its NP-hardness and present a mixed integer linear programming framework with two cutting-plane approaches. Finally, we experimentally compare these approaches on simulated instances of various sizes.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Boscolo, R., Sabatti, C., Liao, J., Roychowdhury, V.: A Generalized Framework for Network Component Analysis. IEEE Trans. Comp. Biol. Bioinf. 2, 289–301 (2005)
Brunet, J.P., Tamayo, P., Golub, T.R., Mesirov, J.P.: Metagenes and Molecular Pattern Discovery Using Matrix Factorization. PNAS 101, 4164–4169 (2004)
Garey, M., Johnson, D.: Computers and Intractability: A Guide to the Theory of NP-Completeness. W.H. Freeman and Company, New York (1979)
Hopcroft, J., Karp, R.: An n 5 / 2 Algorithm for Maximum Matchings in Bipartite Graphs. SIAM J. Comput. 2, 225–231 (1973)
Lee, D., Seung, H.: Learning the Parts of Objects by Non-negative Matrix Factorization. Nature 401, 788–791 (1999)
Liao, J., Boscolo, R., Yang, Y.L., Tran, L., Sabatti, C., Roychowdhury, V.: Network Component Analysis: Reconstruction of Regulatory Signals in Biological Systems. PNAS 100, 15522–15527 (2003)
Lin, C.J.: Projected Gradient Methods for Non-Negative Matrix Factorization. Neural Computation 19, 2756–2779 (2007)
Lovasz, L., Plummer, M.: Matching Theory. North-Holland, Amsterdam (1986)
Murota, K.: Matrices and Matroids for Systems Analysis. Springer, Heidelberg (2000)
Narasimhan, S., Rengaswamy, R., Vadigepalli, R.: Structural Properties of Gene Regulatory Networks: Definitions and Connections. IEEE Trans. Comp. Biol. Bioinf. (accepted, 2007)
Shahnaz, F., Berry, M., Pauca, V., Plemmons, R.: Document Clustering Using Nonnegative Matrix Factorization. Inf. Proc. & Manag. 42, 373–386 (2006)
Wolsey, L.: Integer Programming. Wiley Interscience, Chichester (1998)
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 2008 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Fritzilas, E., Rios-Solis, Y.A., Rahmann, S. (2008). Structural Identifiability in Low-Rank Matrix Factorization. In: Hu, X., Wang, J. (eds) Computing and Combinatorics. COCOON 2008. Lecture Notes in Computer Science, vol 5092. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-69733-6_15
Download citation
DOI: https://doi.org/10.1007/978-3-540-69733-6_15
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-69732-9
Online ISBN: 978-3-540-69733-6
eBook Packages: Computer ScienceComputer Science (R0)